Digital Innovation Labs

You'll learn: Design a repeatable AI-assisted task and check the quality of its results.

You'll apply: Turn sample project notes into a draft update, test different inputs, and improve the instructions.

You'll build: A reusable prompt-and-workflow blueprint for one recurring professional task.

You'll leave with: Your prompt, input checklist, test results, and a review checklist you can reuse.

Named tools by route: Featured: ChatGPT. Low-code: Langfuse prompt versioning. Optional code: Promptfoo testing.

Ways to participate: No-code: adapt prompts and review results. Low-code: structure output fields. Basic-code: optionally add a simple validation check.

You'll learn: Identify and prioritize automation opportunities based on value, effort, and risk.

You'll apply: Map a request process, examine where work stalls, and compare possible improvements.

You'll build: A current-state workflow map and prioritized automation opportunity brief.

You'll leave with: A justified first automation candidate and a sketch of how the improved process could work.

Named tools by route: No-code: Zapier, Make, Scribe, and bpmn.io process mapping. Low-code option: Activepieces.

Ways to participate: No-code: map and trace the process. Low-code: configure a rule or data mapping. Basic-code: optionally test the same rule in a short script.

You'll learn: Design an AI workflow and decide when a fixed sequence or an AI agent fits the task.

You'll apply: Trace a sample policy-based drafting task through normal, missing-information, and out-of-scope situations, with sending simulated.

You'll build: An AI workflow architecture map with human review and escalation points.

You'll leave with: Clear input and output requirements, action limits, and examples of how the workflow handles problems.

Named tools by route: Featured: Gemini. No-code: Tines workflow design. Low-code: Dify. Optional code: Google Agent Development Kit (ADK).

Ways to participate: No-code: map and walk through the workflow. Low-code: configure a decision branch. Basic-code: optionally adapt a small routing script.

You'll learn: Frame a user problem and choose the smallest prototype that can test your idea.

You'll apply: Adapt a starter intake interaction using sample data, observe someone completing a task, and revise it using their feedback.

You'll build: A focused, testable screen flow or simple working prototype.

You'll leave with: Your prototype, test findings, and a focused experiment plan.

Named tools by route: Featured: Replit. No-code screen options: Google Stitch and Magic Patterns. Optional code: Streamlit (Python).

Ways to participate: No-code: build a guided screen or interaction. Low-code: configure fields and validation. Basic-code: optionally adapt a small interactive app.

You'll learn: Structure professional knowledge so it supports useful tasks and decisions.

You'll apply: Organize sample meeting notes, connect decisions to their sources, and test whether someone can find the information they need.

You'll build: A redesigned knowledge workflow and a map of where AI assistance fits.

You'll leave with: A sample knowledge collection, a guide to its fields, and practical review rules.

Named tools by route: No-code: Notion, Tana, Capacities, and the hosted Khoj app. Low-code: structured fields and retrieval rules in Notion or Tana.

Ways to participate: No-code: organize and link information. Low-code: define fields and retrieval rules. Basic-code: optionally check missing information with a short script.

You'll learn: Connect systems through events and design useful responses when something goes wrong.

You'll apply: Use sample events in a sandbox workflow to map data, test incomplete and repeated messages, and trace a simulated failure.

You'll build: An integration blueprint with success, failure, and exception paths.

You'll leave with: A description of the data exchanged, case results, and recovery steps, identifying what you tested and what remains a design.

Named tools by route: No-code: Make, with webhook.site for inspecting events. Low-code: n8n, with Hookdeck for webhook handling. Optional basic-code: Python.

Ways to participate: No-code: adapt and test a visual workflow in Make. Low-code: configure data transformations and conditions in n8n. Basic-code: optionally edit a small event-processing script in Python.

You'll learn: Design human review that lets an authorized person pause, reject, or approve a specific draft before the next step.

You'll apply: Test approval, rejection, timeout, and changed-draft situations, including when a new approval is needed. Keep the final action simulated.

You'll build: A human-review, approval, and escalation control map.

You'll leave with: Review requirements, rules for who can approve which draft, and a record of the situations you tested.

Named tools by route: Featured: Copilot Studio. No-code: Tines. Low-code: gotoHuman review steps. Optional code: LangGraph.

Ways to participate: No-code: map and test review steps. Low-code: configure when work can resume. Basic-code: optionally test decision states in a short script.

You'll learn: Turn an AI idea into a testable hypothesis with clear measures of success.

You'll apply: Adapt a small working AI request-routing starter and compare it with a manual approach using the same sample cases.

You'll build: A working AI prototype for one bounded task and an assumption-and-test card.

You'll leave with: Scored sample results, observations about failures, and a decision about your next experiment.

Named tools by route: No-code: Gumloop. Low-code: Langflow and PostHog experiment measurement. Optional code: Python components in Langflow.

Ways to participate: No-code: adapt a visual prototype and score results. Low-code: configure categories and rules. Basic-code: optionally adapt prototype checks or scoring.

You'll learn: Evaluate sources, challenge assumptions, and connect conclusions to supporting evidence.

You'll apply: Compare options, inspect original source passages, and revise claims that the evidence does not support.

You'll build: A source-backed decision brief and an evidence-and-assumptions matrix.

You'll leave with: A recommendation with traceable sources, stated limitations, and clear reasoning.

Named tools by route: No-code: NotebookLM, Gemini, ChatGPT, Elicit, and Zotero. Optional structured-data activities use the same evidence set.

Ways to participate: No-code: verify sources and write the brief. Optional low-code or basic-code: structure evidence records or check for missing source information.

You'll learn: Design an operational system that people can use, monitor, and maintain.

You'll apply: Adapt one sandbox intake-to-review workflow, trace its records and status, and test a simulated failed handoff or duplicate request.

You'll build: A blueprint connecting data, interfaces, automations, and system connections.

You'll leave with: A data model, workflow stages, ownership and access rules, and recovery notes.

Named tools by route: No-code: Airtable, Make, and Baserow. Low-code: Retool and Appsmith. Optional code: API and validation rules.

Ways to participate: No-code: adapt a form and operator view. Low-code: configure a query or connection. Basic-code: optionally add a record-validation or workflow rule.

You'll learn: Evaluate how a document-based AI system retrieves evidence and uses it to answer questions.

You'll apply: Test a prepared document-based assistant, compare retrieved passages with its answers, and distinguish missing evidence from unsupported answers.

You'll build: A RAG solution design, evaluation set, and failure checklist.

You'll leave with: Test questions, expected evidence, observed results, and a documented improvement.

Named tools by route: No-code: AnythingLLM. Low-code: n8n with a retrieval component such as Qdrant, Weaviate, or Supabase. Optional code: Chroma and Ragas.

Ways to participate: No-code: test answers and inspect sources. Low-code: adjust retrieval settings. Basic-code: optionally adapt how documents are divided, retrieved, or tested.

You'll learn: Evaluate tool choices against the problem, cost, reliability, portability, and maintenance needs.

You'll apply: Use a starter solution or prior lab artifact to compare options for one component, demonstrate a working path and a limitation, and improve your design after feedback.

You'll build: A bounded solution demonstration and an architecture decision record.

You'll leave with: A component map, the reasons behind your choices, and a next-step validation plan.

Named tools by route: No-code architecture: IcePanel. Optional code: Log4brains, Ollama, and Open WebUI. Choose commercial or open-source components to fit your solution.

Ways to participate: No-code: demonstrate a visual solution and explain the choices. Low-code: adapt a component or connection. Basic-code: optionally modify a small adapter or validation rule.